How to make a screenshot with visible mouse cursor

It turns out amazingly uneasy to make a screenshot on which the mouse cursor (pointer) is visible. No one of the typical Linux/Ubuntu screen capture programs does that. I even saw suggestions to run the application inside vmware or virtualbox to make such a screenshot!

There is a better alternative.

Instead of making a screenshot, one should record the short video animation. Those record cursors properly. Then the static image can be extracted from the video.

Recording the video

Run gtk-recordMyDesktop (it is packaged, apt-get install gtk-recordMyDesktop), configure it to grab the appropriate part of the screen (in most cases just use the Select Window button), start recording, make sure the situation which you would like to shot lasts for at least a moment, stop recording (left click on the grey rectangle icon in the panel to which gtk-recordMyDesktop minimized).

Your animation will be saved to ~/out.ogg (or ~/out.ogg.1, ~/out.ogg.2, ... if such file was already present).

In newer recordMyDesktop versions it is ~/out.ogv. Whichever extension is used, this is the Theora-encoded video.

Extracting the image from the video

The only remaining thing is to extract the individual image (frame) from the video.

The simple way to achieve this is to open the video in vlc (also packaged, apt-get install vlc):

$ vlc ~/out.ogg

play the video, pause it on the correct frame, and pick Video/Snapshot from the menu. VLC will save the current frame as something like ~/.vlc/vlcsnap-533222343.png. Rename and move this file, remove the video, job done.

Alternatively, one can convert the whole video to the images sequence. To do this, first convert from Theora to AVI:

$ mencoder ~/out.ogg -o out.avi \
       -ovc lavc -lavcopts vcodec=msmpeg4v2:vbitrate=800

(options do not really matter, one just need to achieve some format readable for ffmpeg), and then issue the following command:

$ mkdir img
$ ffmpeg -qscale 1 -i out.avi -an -r 2 img/frame%05d.png

and all frames of the video will be saved as img/frame00001.png, img/frame00002.png, ...

Note the -r option, it specifies how many frames a second are to be saved. I suggest 2 (making a shot every half a second), it should suffice to find the image needed without creating hundreds of images. If not, try increasing it.

Extra note

This technique can be sometimes fairly effective in case one needs to make a lot of screenshots of some working application. It takes less time to record and unpack the video, then pick appropriate images from the saved frames, then to make every screenshot separately.

It rocks when one need to capture the items uneasy to schedule (like different floating popups, multilevel menus etc).